Skip to content

Latest commit

 

History

History
416 lines (309 loc) · 18.1 KB

execution.rst

File metadata and controls

416 lines (309 loc) · 18.1 KB
Raw

Execution

The MaNGA DAP was originally developed to analyze data from the SDSS-IV/MaNGA Survey, provided reduction products from the MaNGA data-reduction pipeline (DRP). However, it is now easy to apply the MaNGA DAP to any IFU datacube.

Below, we describe how to use the MaNGA DAP to analyze MaNGA datacubes, as well as how to execute the survey-level batch mode that analyzes all the MaNGA datacubes within a given directory structure.

To analyze non-MaNGA datacubes, see both fitdatacube and the description of the MaNGA DAP plan. Also, for an example of how to fit single spectra using MaNGA DAP core algorithms, see fitonespec.

Input files

DAP AnalysisPlan

The DAP uses a toml file to define parameters used by its modules to analyze the provided datacube. This is described in detail by plan.

Default AnalysisPlan

If executed without an AnalysisPlan parameter file, the command-line execution of the DAP will use a default plan; see mangadap.config.analysisplan.AnalysisPlan.default and default_plan.toml.

The DAP Datacube Configuration File

The DAP uses a configuration (ini) file to set the datacube to be analyzed and provides some relevant metadata. These configuration files are generated at the survey-level by ~mangadap.survey.drpcomplete.write_config. However, we also provide the write_dap_config script that will generate the relevant configuration files if you have the DRPall or DRPComplete file. As with all the DAP scripts, you can use the -h command-line option to get the usage:

To construct the configuration file for datacube 7815-3702, run:

write_dap_config 7815 3702 mangadap-7815-3702.ini -a drpall-v3_0_1.fits

to produce:

# Auto-generated configuration file
# Fri 28 Feb 2020 16:57:19

[default]
drpver
redux_path
directory_path
plate = 7815
ifu = 3702
log = True
sres_ext
sres_fill
covar_ext
z = 2.9382300e-02
vdisp
ell = 1.1084400e-01
pa = 1.6324500e+02
reff = 3.7749500e+00

Use the relevant keywords to change the paths or the extensions used for the spectral resolution and spatial correlation matrix (e.g., GCORREL).

Note

The metadata included in the configuration file is used by the DAP during the analysis; however, not all of these metadata parameters are explicitly required. See how the data is checked/used in mangadap.datacube.datacube.DataCube.populate_metadata.

DAP command-line script

The main DAP script is manga_dap, which is a simple wrapper of ~mangadap.survey.manga_dap.manga_dap. With the DAP installed, you can call the script directly from the command line:

The DAP allows you to define your own datacube class, as long as it is derived from ~mangadap.datacube.datacube.DataCube. You can then specify that your data should be instantiated with that derived class using the cube_module argument; this defaults to mangadap.datacube.datacube.MaNGADataCube expecting that you're analyzing a MaNGA datacube.

When running the DAP on a MaNGA datacube, you have to provide a configuration file; however, for derived classes, you may be able to fully instantiate the relevant data just using the datacube file, which is why we've provided the -f option.

Note that the analysis plan file is an optional argument. If it is not given, the DAP will use the execution-analysis-plan-default. As discussed in plan, you can also provide a derived class for the analysis plan, which should basically be provide fine-tuned control over the output directory structure. If you're analyzing a non-MaNGA datacube, you will likely be fine just using the analysis plan base class. However, because manga_dap defaults to using the MaNGA specific plan class, you'll need to specify the base class (mangadap.config.analysisplan.AnalysisPlan) when executing the code.

To run the DAP on a single datacube using the default analysis plan, and assuming you have the DRPall file and the relevant LOGCUBE and LOGRSS files (see the warning below) in the current directory, you could execute the DAP as follows:

write_dap_config 7815 3702 mangadap-7815-3702.ini -a drpall-v3_1_1.fits --directory_path .
manga_dap -c mangadap-7815-3702.ini -vv --log mangadap-7815-3702.log -d . -o dap_output

This will analyze the datacube for observation 7815-3702 using the default analysis plan, with verbose output and a log written to mangadap-7815-3702.log, and with the root directory for all the DAP output (except for the log) set to dap_output. These commands can be successfully run in the $MANGADAP_DIR/mangadap/data/remote directory if you've run the download_test_data.py script.

Warning

When running the DAP on a MaNGA datacube, you should have both the DRP LOGRSS and LOGCUBE files in the directory_path if you want to account for the spatialcovariance! If the LOGRSS files are not present, the DAP will throw a warning and continue, which means that the warning can get buried among all the other messages and likely missed.

Programmatic execution

The manga_dap executable script is simple, in that it only (1) reads the datacube, (2) sets the analysis plan, and (3) executes the main DAP analysis wrapper function, mangadap.survey.manga_dap.manga_dap. It is straight-forward then to construct a script that executes the DAP programmatically for many datacubes.

Batch execution using automatically generated scripts

Note

Version 4.x of the DAP has never been used for a batch execution/analysis of the MaNGA data. The code should work properly. If it doesn't, please Submit an issue.

The survey-level execution of the DAP uses the rundap script, which is a simple wrapper of ~mangadap.survey.rundap.rundap. This script

  • sets up the DAP output directory structure
  • either confirms that a provided list of datacubes to analyze exist on disk or trolls the DRP directory structure to find all or some subset of available datacubes to analyze
  • creates execution-config for each plateifu to be analyzed,
  • creates a script file for each plateifu that can be sourced to execute the DAP and the associated QA plots,
  • creates scripts that execute the plate-level QA plots,
  • creates scripts that build the DAPall file and its QA plots, and
  • submits the scripts to the Utah cluster.

The last step uses an SDSS python package called pbs, which isn't required for the more general-purpose use of the rundap script discussed here. With the DAP installed, you can call the script directly from the command line:

If a execution-analysis-plan is not provided, the scripts will use the execution-analysis-plan-default.

An example call of this script that will only construct scripts for the analysis of observation 7443-12701 using the default AnalysisPlan is:

rundap --platelist 7443 --ifudesignlist 12701 --redux_path /path/with/drp/output/drpver --analysis_path /path/for/dap/output/drpver/dapver -vv --log

In this call, I've specified that the DRP data is in /path/with/drp/output/ and that the DAP output should be placed in /path/for/dap/output/ instead of using the default datamodel-directory-structure.

The script file this call produces is written to /path/for/dap/output/log/[time]/7495/12704/mangadap-7495-12704-LOGCUBE, where [time] is a time stamp of when rundap was executed. (If you execute rundap multiple times, it will create new directories using new time stamps each time.) The lines of the script file for each plate-ifu:

  • touches the *.started file
  • executes manga_dap
  • executes a series of QA plotting scripts
  • touches the *.done file

The example script generated by the above command would look something like this:

# Auto-generated batch file
# Wed 06 Apr 2022 15:15:26

touch /path/for/dap/output/drpver/dapver/log/06Apr2022T22.15.26UTC/7443/12701/mangadap-7443-12701-LOGCUBE.started

OMP_NUM_THREADS=1 manga_dap -c /path/for/dap/output/drpver/dapver/common/7443/12701/mangadap-7443-12701-LOGCUBE.ini -o /path/for/dap/output/drpver/dapver --log /path/for/dap/output/drpver/dapver/log/06Apr2022T22.15.26UTC/7443/12701/mangadap-7443-12701-LOGCUBE.log -vv

OMP_NUM_THREADS=1 dap_ppxffit_qa -c /path/for/dap/output/drpver/dapver/common/7443/12701/mangadap-7443-12701-LOGCUBE.ini -o /path/for/dap/output/drpver/dapver -b 2.5

OMP_NUM_THREADS=1 spotcheck_dap_maps -c /path/for/dap/output/drpver/dapver/common/7443/12701/mangadap-7443-12701-LOGCUBE.ini -o /path/for/dap/output/drpver/dapver -b 2.5

OMP_NUM_THREADS=1 dap_fit_residuals -c /path/for/dap/output/drpver/dapver/common/7443/12701/mangadap-7443-12701-LOGCUBE.ini -o /path/for/dap/output/drpver/dapver -b 2.5

touch /path/for/dap/output/drpver/dapver/log/06Apr2022T22.15.26UTC/7443/12701/mangadap-7443-12701-LOGCUBE.done

To execute the script, you would then run:

source /path/for/dap/output/drpver/dapver/log/06Apr2022T22.15.26UTC/7443/12701/mangadap-7443-12701-LOGCUBE

The rundap script allows you to construct scripts for all datacubes it can find on disk, all IFUs on a given plate, all combinations of a set of plate and IFU numbers, or for a specified list of plateifu IDs.

Note

The rundap script constructs the ~mangadap.survey.drpcomplete.DRPComplete object and writes its associated fits file; see metadatamodel-drpcomplete. The data compiled into this database is pulled from the DRPall file, but some corrections may be applied to the NSA redshift or photometry; see, e.g., metadatamodel-redshift-fix.

To write the post-processing scripts, execute rundap with the --post and --post_plots options. This produces two additional types of scripts:

  • Scripts to produce QA plots for all IFUs on a given plate. This file is written to, e.g., /path/for/dap/output/drpver/dapver/log/06Apr2022T22.15.26UTC/7443/7443_fitqa and looks like this:

    # Auto-generated batch file
# Thu 07 Apr 2022 13:33:27

touch /path/for/dap/output/drpver/dapver/log/07Apr2022T20.33.27UTC/10001/10001_fitqa.started

OMP_NUM_THREADS=1 dap_plate_fit_qa 10001 --analysis_path /path/for/dap/output/drpver/dapver --plan_file /path/for/dap/output/drpver/dapver/log/07Apr2022T20.33.27UTC/plan.toml

touch /path/for/dap/output/drpver/dapver/log/07Apr2022T20.33.27UTC/10001/10001_fitqa.done

  • A script that builds the metadatamodel-dapall and writes its QA plots. This file is written to, e.g., /path/for/dap/output/drpver/dapver/log/06Apr2022T22.15.26UTC/build_dapall and looks like this:

    # Auto-generated batch file
# Thu 07 Apr 2022 13:33:27

touch /path/for/dap/output/drpver/dapver/log/07Apr2022T20.33.27UTC/build_dapall.started

OMP_NUM_THREADS=1 construct_dapall --drpver v3_1_1 -r /path/with/drp/output/drpver --dapver 3.1.0 -a /path/for/dap/output/drpver/dapver --plan_file /path/for/dap/output/drpver/dapver/log/07Apr2022T20.33.27UTC/plan.toml -vv

OMP_NUM_THREADS=1 dapall_qa --drpver v3_1_1 --redux_path /path/with/drp/output/drpver --dapver 3.1.0 --analysis_path /path/for/dap/output/drpver/dapver --plan_file /path/for/dap/output/drpver/dapver/log/07Apr2022T20.33.27UTC/plan.toml

touch /path/for/dap/output/drpver/dapver/log/07Apr2022T20.33.27UTC/build_dapall.done

In the automated run of the DAP, any entry in the metadatamodel-drpcomplete that meets the criteria set by mangadap.survey.drpcomplete.DRPComplete.can_analyze will be analyzed. Currently the relevant criteria are:

  • MANGAID != NULL
  • MANGA_TARGET1 > 0 | MANGA_TARGET3 > 0
  • VEL > -500

An important consequence of the selection above is that any targets without a provided redshift will not be analyzed by the DAP, unless it has replacement redshift in the metadatamodel-redshift-fix. Ancillary targets not analyzed by the DAP are likely because a redshift was not available.

Execution Recovery

If the code faults, all of the modules completed up until the fault occurred should have created a "reference" file that will effectively allow the code to pick up where it left off. The mechanism used to determine if it can do this relies simply on the existence of the expected output file. This means that if you change one of the parameters in your input plan file without changing the keyword identifier for that module parameter set, the code may read in the existing file and keep going without incorporating your parameter change.

If you're testing the performance with different parameter values, either perform the testing with different output directories, always remember to change the keyword for the relevant module parameter set, set the overwrite keyword for the relevant module (and each subsequent module!) to True, or just nuke the directory and start again.

Quality Assessment Plots

The survey-level execution of the MaNGA DAP constructs automatically generated plots that provide spotchecks of the performance of the DAP; see qualityassessment. The main QA plots for the analysis of a single datacube have a similar calling sequence as the main execution-mangadap. The three main scripts are spotcheck_dap_maps, dap_ppxffit_qa, and dap_fit_residuals. Following the example execution of the execution-mangadap above, these can be created using the following subsequent calls:

spotcheck_dap_maps -c mangadap-7815-3702.ini -vv --log mangadap-7815-3702.log -d . -o dap_output
dap_ppxffit_qa -c mangadap-7815-3702.ini -vv --log mangadap-7815-3702.log -d . -o dap_output
dap_fit_residuals -c mangadap-7815-3702.ini -vv --log mangadap-7815-3702.log -d . -o dap_output

Local Environment Setup for Survey-Level MaNGA Analysis

The DAP uses environmental variables to define the paths to specific data and other repositories, when executed for MaNGA data. If these are not defined, default values will be used; see the initialization of the mangadap.config module. The relevant environmental variables, their default, and their usage are provided below.

Variable Default Comments
MANGADRP_VER v3_1_1 (i.e., MPL-11) Version of the DRP, used for path construction
MANGA_SPECTRO_REDUX $HOME/MaNGA/redux Root path for the reduced data
MANGADAP_VER mangadap.__version__ Version of the DAP, used for path construction
MANGA_SPECTRO_ANALYSIS $HOME/MaNGA/analysis Root path for the analysis data

These environmental variables can be added to, e.g., your .bash_profile file in your home directory or be included in a script that is sourced when you want to run the DAP. The lines added to your .bash_profile file could look something like this:

export MANGA_SPECTRO_REDUX=/path/with/drp/output
export MANGADRP_VER=v3_1_1

export MANGA_SPECTRO_ANALYSIS=/path/for/dap/output
export MANGADAP_VER=3.1.0

Note

  • Importantly, note that $MANGADAP_VER is only used to set the path names, not to select the specific version of the DAP that should be used. The version of the DAP used is always the one installed by your python environment.
  • The DAP checks that these variables are defined every time it is imported.
  • Some of these same variables are defined by Marvin. It is possible to have both Marvin and the DAP point to the same directory, but beware that this may mean that some of the files get overwritten!
  • Two additional variables ($MANGACORE_VER and $MANGACORE_DIR) are used in a specific mode of survey-level execution of the DAP. However, this is a niche usage mode and is effectively never used. See execution-rundap.
  • The DAP expects to find the DRP LOGCUBE and LOGRSS files in the directory $MANGA_SPECTRO_REDUX/$MANGADRP_VER/[PLATE]/stack, where [PLATE] is the desired plate number. The LOGRSS files are required if you want to properly account for spatialcovariance. This path can be altered when executing the DAP.
  • The DAP expects to find/write data to $MANGA_SPECTRO_ANALYSIS/$MANGADRP_VER/$MANGADAP_VER. This path can be altered when executing the DAP, but the subdirectory structure used by the DAP to organize its outputs within this root directory cannot currently be changed.